Yahoo 知識+ 將於 2021 年 5 月 4 日 (美國東岸時間) 停止服務,而 Yahoo 知識+ 網站現已轉為僅限瀏覽模式。其他 Yahoo 資產或服務,或你的 Yahoo 帳戶將不會有任何變更。你可以在此服務中心網頁進一步了解 Yahoo 知識+ 停止服務的事宜,以及了解如何下載你的資料。
Bounded Variation and Interpolation on Infinitely Many Nodes?
I thought of a fun little exercise in measure theory that turned out to be more interesting than originally anticipated, and I wanted to share.
Prove or provide a counterexample to the following 'conjecture':
Suppose (a_n) is a monotone (increasing) sequence of real numbers contained in the interval (0,1), and suppose that Lim_{n -> infinity} a_n = 1.
Also, suppose (b_n) is a bounded sequence of real numbers.
Then there exists a function f in BV([0,1]) such that f'(a_n) = b_n for all n >= 1, if and only if
Lim_{n -> infinity} b_n
exists.
I'm not doing this for a class. I just thought it was an interesting question. I have a complete proof of the backwards direction (I broke it into a few simple cases and one not-so-simple case).
The more I look at it, the more I doubt the forward direction is even true. If, perhaps, we assume that
Sum_{n=1 to infinity} |a_{n+1} - a_n| < infinity
then maybe a quick counterexample will rear its ugly head.
Do you mean
b_n = f'(a_n) = sin(1-1/(n+1))?
Because this does have a limit, namely sin(1). Otherwise, would you mind elaborating on your answer a bit here?
Admittedly, the Sine Integral function was a candidate I considered briefly, but I wasn't able to get anything particularly useful from this.
I see what you did. It wasn't that I was making the wrong substitution. I was simply misinterpreting what you had written originally. That does indeed seem to be a counterexample to the forward direction. I was originally thinking one might be able to design a counterexample giving b_n = (-1)^n (which a simple modification of your example would give). Thanks!
1 個解答
- 匿名8 年前最愛解答
What about:
f(x) = int_{t=0}^{x} sin(1/(t-1)) dt, if x in [0, 1).
f(1) = 0.
Then f'(x) = sin(1/(x-1)) for x in (0,1), and:
int_{x=0}^{x=1} |f'(x)|dx <= 1.
So f(x) has bounded variation.
Now define the increasing sequence:
a_n = 1 - 1/(n+1) for n in {1, 2, 3, …}
Define b_n by:
b_n = f'(a_n) = sin(-(n+1))
Then {b_n} is a bounded sequence with no limit.
***
@Nick: To get f'(a_n), I am substituting "1 - 1/(n+1)" for "x" in f'(x).
This indeed leads to sin(-(n+1)). You seem to be incorrectly
substituting. You are accidentally substituting "1-1/(n+1)" for "1/(x-1)."